Phases of a bilayer Fermi gas Please share

advertisement
Phases of a bilayer Fermi gas
The MIT Faculty has made this article openly available. Please share
how this access benefits you. Your story matters.
Citation
Nishida, Yusuke, "Phases of a bilayer Fermi gas." Physical
Review A 82.1(2010): 011605(R). © 2010 The American
Physical Society
As Published
http://dx.doi.org/10.1103/PhysRevA.82.011605
Publisher
American Physical Society
Version
Final published version
Accessed
Thu May 26 17:31:26 EDT 2016
Citable Link
http://hdl.handle.net/1721.1/60576
Terms of Use
Article is made available in accordance with the publisher's policy
and may be subject to US copyright law. Please refer to the
publisher's site for terms of use.
Detailed Terms
RAPID COMMUNICATIONS
Selected for a Viewpoint in Physics
PHYSICAL REVIEW A 82, 011605(R) (2010)
Phases of a bilayer Fermi gas
Yusuke Nishida
Center for Theoretical Physics, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA
(Received 6 July 2009; published 12 July 2010)
We investigate a two-species Fermi gas in which one species is confined in two parallel layers and interacts
with the other species in the three-dimensional space by a tunable short-range interaction. Based on the controlled
weak coupling analysis and the exact three-body calculation, we show that the system has a rich phase diagram in
the plane of the effective scattering length and the layer separation. Resulting phases include an interlayer s-wave
pairing, an intralayer p-wave pairing, a dimer Bose-Einstein condensation, and a Fermi gas of stable Efimov-like
trimers. Our system provides a widely applicable scheme to induce long-range interlayer correlations in ultracold
atoms.
DOI: 10.1103/PhysRevA.82.011605
PACS number(s): 03.75.Ss, 05.30.Fk, 74.20.Rp, 74.78.−w
I. INTRODUCTION
One of the central themes in ultracold atoms is to provide
highly tunable model systems for other subfields in physics.
A prominent example is the realization of ultracold twodimensional atomic gases which have enabled detailed microscopic studies of Berezinskii-Kosterlitz-Thouless physics
relevant to a wide variety of two-dimensional phenomena
[1–3]. In addition to simple two-dimensional systems with
single layers, what have attracted considerable attention in
condensed matter physics are bilayer or multilayer systems.
Here, extra degrees of freedom generated by layers and the
long-range Coulomb interaction between them are expected
to lead to intriguing physics such as an interlayer exciton
condensation in bilayer semiconductors [4,5], quantum Hall
bilayers [6], and bilayer graphenes [7].
In ultracold atoms, the analogous multilayer geometry can
be created by confining atoms by a strong optical lattice
in one direction. However, since the long-range Coulomb
interaction is absent in neutral atoms, separated layers are
simply decoupled without interlayer tunneling. A scheme that
we propose in this article to realize the long-range interlayer
correlation in ultracold atoms is to use a mixture of two atomic
species A and B and confine only A atoms in the multilayer
geometry with keeping B atoms in the three-dimensional
space. Here the correlation between A atoms confined in
different layers can be induced through the interaction with
the background B atoms that are free to propagate from
layer to layer. The advantage of this scheme is that one can
tune the A-B interaction by interspecies Feshbach resonances
and thus rich phenomena are expected to occur as a function
of the interaction strength. Furthermore, such a system may
be thought of as a tunable model system for a slab phase
of nuclear matter by identifying A (B) atoms with protons
(neutrons) [8,9].
In this article, we give detailed analyses in the case of a
bilayer Fermi gas that can be realized by using a Fermi-Fermi
mixture of, for example, 6 Li and 40 K [10–12] and confining
only one species in two parallel layers. Based on the controlled
weak coupling analysis and the exact three-body calculation,
we show that the system exhibits at least four distinct quantum
phases as summarized in Fig. 1. In the weak coupling region,
1050-2947/2010/82(1)/011605(4)
011605-1
an effective attraction between two A atoms mediated by B
atoms leads to superfluidity due to an interlayer s-wave pairing
(lower left phase) or an intralayer p-wave pairing (upper left
phase) depending on the layer separation. On the other hand,
in the strong coupling region, an A atom captures a B atom to
form a tightly bound molecule and the ground state becomes
a dimer Bose-Einstein condensation (right phase). Finally,
when the A-B interaction is close to the resonance, we find
that two A atoms confined in different layers with one B
atom form a three-body bound state leading to a Fermi gas of
trimers in a dilute system (lower middle phase). Remarkably,
there is an infinite number of such three-body bound levels
exactly at the resonance resembling the Efimov effect in three
dimensions [13].
The system under consideration is described by the action
(here and below h̄ = 1 and kB = 1):
∇2
†
S=
dtd xψAi (t,x) i∂t + x + µA ψAi (t,x)
2mA
i=1,2
∇x2 + ∇z2
†
+ µB
+ dtd xdzψB (t,x,z) i∂t +
2mB
kF d
O(1)
∞
intralayer p-wave SF
dimer
BEC
O(1)
interlayer
s-wave SF
trimer FG
0
−∞
0
1
+∞ aeff kF
FIG. 1. Conjectured phase diagram at zero temperature in the
plane of the effective scattering length (aeff kF )−1 and the layer
separation kF d. Here, kF ∼ kFA ∼ kFB is assumed.
©2010 The American Physical Society
RAPID COMMUNICATIONS
YUSUKE NISHIDA
PHYSICAL REVIEW A 82, 011605(R) (2010)
p, i
kFB d
q, i
2.0
+ O(a3eff )
−i Ṽij (|p − q|) =
−p, j
1.5
0
1
V 12 V 11
−q, j
1.0
FIG. 2. Effective interaction between two A atoms (solid line)
induced by the interaction with B atoms (dotted line) in the weak
coupling limit. Here, i,j = 1,2 are layer indices.
× ψB (t,x,z) + g0
†
†
i=1,2
(1)
Here, ψAi (t,x) with x = (x,y) is a spinless fermionic field for
A atoms confined in two parallel layers that are labeled by
the index i = 1,2 and located at z = zi . ψB (t,x,z) is another
spinless fermionic field for B atoms in the three-dimensional
space. The last term in the action describes the short-range
A-B interaction. g0 is a cutoff () dependent bare coupling,
which can be eliminated by introducing the physical
parameter,
√
mB mAB
1
the effective scattering length aeff , through g0 − 2π =
√
m m
B AB
with mAB ≡ mA mB /(mA + mB ) being the reduced
− 2πa
eff
mass [14,15]. aeff is arbitrarily tunable by means of the
interspecies Feshbach resonance and the limit aeff → −(+)0
corresponds to the weak (strong) attraction between A and
B atoms. Throughout this article, the interlayer tunneling is
assumed to be negligible and Fermi momenta of A and B
atoms are defined through their densities; kFA ≡ (4π nA )1/2
and kFB ≡ (6π 2 nB )1/3 .
II. WEAK COUPLING LIMIT
In order to elucidate phases appearing in our system, we
start with the weak coupling limit aeff → −0 in which a controlled perturbative analysis is possible. Unlike the ordinary
BCS-BEC crossover purely in two or three dimensions, the
A-B pairing at weak coupling does not take place in our
mixed dimensional system because of the absence of the
Cooper instability [15]. What can happen instead are pairings
between A atoms using the effective attraction induced by
the interaction with the Fermi sea of B atoms. To the leading
order in aeff , the back-to-back scattering of A atoms in layers
i and j is described by the Feynman diagram in Fig. 2. The
resulting induced interaction Ṽij (| p − q|) has rather a simple
form in the real space:
Vij (|x|) =
2
aeff
2kFB rij cos(2kFB rij ) − sin(2kFB rij )
,
mAB
4π rij4
(2)
where rij = x 2 + (zi − zj )2 is the distance between the two
A atoms. The oscillatory decaying factor is well known in
the Ruderman-Kittel-Kasuya-Yosida interaction [16–18] and
leads to the following physical interpretation of Vij (|x|): The
density modulation of background B atoms produced by one
A atom in the layer i mediates the long-range interaction with
the other A atom in the layer j .
0
1
V 12 V 11
0.0
0
dtd xψAi (t,x)ψB (t,x,zi )
× ψB (t,x,zi )ψAi (t,x).
0.5
1
2
3
4
kFA
5 kFB
FIG. 3. (Color online) Critical layer separation kFB d dividing the
interlayer s-wave pairing (Ṽ12(0) < Ṽ11(1) ) from the intralayer p-wave
pairing (Ṽ12(0) > Ṽ11(1) ) as a function of kFA /kFB .
Now the pairings of A atoms are described by the BCS-type
Hamiltonian:
2
†
p
− µA ψ̃Ai ( p)
ψ̃Ai ( p)
Hind =
2mA
p
i
1 †
†
+
ψ̃ ( p)ψ̃Aj (− p)
2 i,j p,q Ai
× Ṽij (| p − q|)ψ̃Aj (−q)ψ̃Ai (q).
(3)
In the mean-field
approximation, the gap function is given by
ij ( p) = 1 q Ṽij (| p − q|)ψ̃Aj (−q)ψ̃Ai (q) and the Fermi
statistics of A atoms implies j i (− p) = −ij ( p). Although
A atoms are spinless fermions, the layer indices i,j = 1,2
make more than one pairing pattern possible. When the
layer separation is large d ≡ |z1 − z2 | → ∞, the interlayer
interaction Vi=j is suppressed while the intralayer interaction
Vi=j is unaffected by d. Therefore, in this limit, intralayer
pairings with 11 ,22 = 0 are favored while their pairing
symmetry has to be p wave [15]. On the other hand, when
the layer separation becomes small d → 0, Vi=j is no longer
suppressed and thus an interlayer pairing with 12 = −21 =
0 is favored because the pairing symmetry can be s wave only
in this singlet channel. Therefore, there has to be a quantum
phase transition as a function of the layer separation.
The quantum phase transition can be located by comparing the energy densities for the interlayer s-wave and
intralayer p-wave pairings. In the weak coupling limit, their
mA µ2A
−
energy densities are, respectively, given by Hind = − 2π
mA |12 |2
4π
and Hind = −
mA µ2A
2π
−
mA |11 |2
8π
−
mA |22 |2
,
8π
where the
(0)
gap functions at | p| = kFA are given by 12 /εFA ∝ e2π/[mA Ṽ12 ]
(1)
and 11 /εFA = 22 /εFA ∝ (p̂x ± i p̂y )e2π/[mA Ṽ11 ] for negative
2
Ṽ (l) . Here, εFA ≡ kFA
/(2mA ) is the Fermi energy and Ṽij(l) ≡
ijπ dϕ
0 π cos(lϕ)Ṽij (| p − q|) with cos ϕ ≡ p̂ · q̂ is the partial
wave projection of the induced interaction in which both
incoming and outgoing momenta are restricted on the Fermi
surface of A atoms; | p| = |q| = kFA . It is the phase with the
smaller energy density and hence the larger pairing gap that is
realized in our system.
The critical layer separation dividing the above two phases
is plotted in Fig. 3 as a function of kFA /kFB , which is
011605-2
RAPID COMMUNICATIONS
PHASES OF A BILAYER FERMI GAS
iT11
=
PHYSICAL REVIEW A 82, 011605(R) (2010)
d 2 mAB E
iT12
0.2
iT12
+
=
0.2
0.4
d
0.6 aeff
0.1
iT11
0.2
FIG. 4. Three-body scattering of two A atoms in different layers
and one B atom. The double line represents the A-B scattering
amplitude. T22 and T21 satisfy the same integral equations as T11
and T12 .
0.3
0.4
0.5
independent of aeff and mA /mB to the leading order in aeff .
(0)
(1)
Below the critical separation where Ṽ12
< Ṽ11
, the interlayer
s-wave pairing appears, while the intralayer p-wave pairing
(0)
(1)
> Ṽ11
. The
appears above the critical separation where Ṽ12
two layers exhibit superfluidity in both phases and its critical
temperature at weak coupling is the same order as the pairing
gap at zero temperature; Tc ∼ |ij |. We note that our interlayer
s-wave pairing bears an analogy to the interlayer exciton
condensation in condensed matter systems [4–7]. Also it is
worthwhile to point out that the intralayer px + ipy -wave
pairing has a potential application to topological quantum computation using vortices with non-Abelian statistics [19,20].
III. STRONG COUPLING LIMIT
We now turn to the strong coupling limit aeff → +0, in
which A atoms confined in layers capture B atoms from the
bulk space to form two-body bound states (dimers) whose
binding energy is given by E2 = − 2m 1 a 2 . The resulting
AB eff
system consists of the dimers localized around layers which
interact weakly with each other and excess B atoms. As long
as the dimer size ∼ aeff is smaller than the layer separation
−1
d and the mean interparticle distance ∼ kFA
, the dimers
behave as two-dimensional bosons and therefore the ground
state becomes a dimer Bose-Einstein condensation in each
layer. Accordingly, the two layers exhibit superfluidity up
to the Berezinskii-Kosterlitz-Thouless temperature given by
2
d
ln−1 (− 380
ln nd aeff
) in the limit aeff → +0 [21],
TBKT → 2πn
M
4π
where M ≡ mA + mB and nd = nA are the dimer’s mass
and density per layer. We note that the dimer Bose-Einstein
condensates in different layers can interact with each other
through the excess B atoms just as in Fig. 2 and Eq. (2).
Although our system is found to exhibit superfluidity in both
weak and strong coupling limits aeff → ∓0, their symmetry
breaking patterns are different. The action in Eq. (1) has
continuous symmetries of U(1)A1 × U(1)A2 × U(1)B × U(1)z
corresponding to particle number conservations of A atoms in
each layer, that of B atoms, and a rotation about z axis. An order
parameter of the interlayer s-wave pairing ij ψAi ψAj = 0
breaks the full symmetries down to U(1)A1−A2 × U(1)B ×
U(1)z , while order parameters of the intralayer px +ipy ↔
↔
wave pairing ψAi ( ∂ x + i ∂ y )ψAi = 0 break them down to
U(1)A1+A2−z × U(1)B . Order parameters of the dimer BoseEinstein condensation are ψA1 ψB ,ψA2 ψB = 0, which
break the full symmetries down to U(1)A1+A2−B × U(1)z .
Therefore, the phases in the weak and strong coupling limits
have to be divided by at least one quantum phase transition as
0.6
FIG. 5. (Color online) Lowest binding energies of trimers for
mass ratios mA /mB = 6.67 (bottom), 1 (middle), and 0.15 (top) as a
function of d/aeff . The dashed and dotted lines are atom-dimer and
dimer-dimer thresholds; E = E2 and 2E2 .
a function of the interaction strength. Below we show that a
novel phase can appear in between when the layer separation
is small.
IV. UNITARITY LIMIT
The fourth phase realized in our system can be elucidated
by studying a three-body problem of two A atoms confined in
different layers interacting with one B atom. Their scattering
process is depicted in Fig. 4 and all the relevant diagrams can
be summed by solving the integral equations for Tij (E; p,q).
Here E is the total energy in the center-of-mass frame and
p (q) is the relative momentum of the incoming (outgoing)
A atom in the layer i (j ) that does not scatter with the B
atom first (last). The T -matrix elements have properties T11 =
T22 and T12 = T21 and can be decomposed into even- and
odd-parity parts under the exchange of one layer index; T± ≡
T11 ± T12 . After the partial wave projection T±(l) (E; p,q) ≡
π dϕ
(l)
0 π cos(lϕ)T± (E; p,q) with cos ϕ ≡ p̂ · q̂, we find T± to
satisfy
T±(l) (E; p,q)
√
= ∓ mB mAB K (l) (E + i0+ ; p,q)
∞
T (l) (E; p,k)K (l) (E + i0+ ; k,q)
∓
dkk ±
,
mB +mAB 2
+− 1
0
k
−
2m
E
−
i0
AB
M
aeff
(4)
where K (l) (E; p,q) is given by
0
π
−d
M
mA
p2 +q 2 +
dϕ
e
cos(lϕ) π
p2 + q 2 +
2mA
M
2mA
pq
M
pq cos ϕ−2mAB E
cos ϕ − 2mAB E
. (5)
The spectrum of three-body bound states (trimers) is
obtained by poles of T±(l) as a function of E. When E
approaches one of the binding energies E3 < − 2mθ(aeffa)2 , we can
AB eff
write T±(l) as T±(l) (E; p,q) → Z±(l) (p,q)/(E − E3 ). By solving
the homogeneous integral equation from Eq. (4) satisfied by
the residue Z±(l) , we find that the trimers exist only in an
odd-parity (−) and s-wave (l = 0) channel. The lowest binding
energy E3(0) as a function of d/aeff is plotted in Fig. 5 for
011605-3
RAPID COMMUNICATIONS
YUSUKE NISHIDA
PHYSICAL REVIEW A 82, 011605(R) (2010)
three different values of the mass ratio mA /mB = 0.15, 1, and
6.67, corresponding to a mixture of A = 6 Li and B = 40 K,
two different internal states of the same atomic species, and
A = 40 K and B = 6 Li. The trimer appears on the negative
side of the effective scattering length |aeff | ∼ d and its binding
energy decreases as d/aeff is increased. When d/aeff → +∞,
the trimer binding energy asymptotically approaches the
atom-dimer threshold E = E2 from below as E3 /E2 − 1 ∝
−
mB +mAB
M
exp(
M d
mA aeff
)
→ +0.
Interestingly, when the A-B interaction is exactly at the twobody resonance |aeff | → ∞, there exists an infinite number of
such trimer states whose spectrum is expressed by the form
2
E3(n) = −e−2πn/s0 2mκAB d 2 for n → ∞ (not shown in Fig. 5).
Such a geometric spectrum at the resonance is well known
as the Efimov effect in three dimensions [13]. Here the
scaling exponent s0 and the so-called Efimov parameter κ can
be determined as (s0 ,κ) = (0.741,0.0311), (0.828,0.0807),
(1.30,0.231) for mA /mB = 0.15, 1, 6.67, respectively. An
important characteristic of our trimer state in contrast to
Efimov trimers in a free space is its stability against the threebody recombination because the two A atoms are spatially
separated.
The existence of stable AAB trimers, which has the same
quantum number as ij ψAi ψAj ψB , can lead to a Fermi gas
of trimers in a dilute system. As long as the trimer size
(2mAB |E3 |)−1/2 ∼ d is smaller than the mean interparticle dis−1
tance ∼kFA
, the trimers behave as two-dimensional fermions
localized around layers and form a Fermi gas with its density
equal to nA . Although the trimer state exists even in the
strong coupling limit aeff → +0, the trimer gas phase cannot
persist there because an addition of another B atom breaks
up the trimer into two dimers when E3 > 2E2 is reached
(see the dimer-dimer threshold in Fig. 5). Therefore, the
trimer Fermi gas is realized only when the A-B interaction
> d. We note that AAB
is close to the resonance |aeff | ∼
e
[1] Z. Hadzibabic et al., Nature (London) 441, 1118 (2006).
[2] V. Schweikhard, S. Tung, and E. A. Cornell, Phys. Rev. Lett. 99,
030401 (2007).
[3] P. Krüger, Z. Hadzibabic, and J. Dalibard, Phys. Rev. Lett. 99,
040402 (2007).
[4] Yu. E. Lozovik and V. I. Yudson, JETP Lett. 22, 274 (1975);
Solid State Commun. 19, 391 (1976); Sov. Phys. JETP 44, 389
(1976).
[5] S. I. Shevchenko, Sov. J. Low Temp. Phys. 2, 251 (1976).
[6] See, e.g., J. P. Eisenstein and A. H. MacDonald, Nature (London)
432, 691 (2004), and references therein.
[7] See, e.g., M. Y. Kharitonov and K. B. Efetov, e-print
arXiv:0903.4445, and references therein.
[8] D. G. Ravenhall, C. J. Pethick, and J. R. Wilson, Phys. Rev. Lett.
50, 2066 (1983).
[9] M. Hashimoto, H. Seki, and M. Yamada, Prog. Theor. Phys. 71,
320 (1984).
trimers with A atoms in the same layer and ABB trimers
are absent for 0.0351 < mA /mB < 6.35 [14]. Once they are
formed, the whole phase diagram would be dominated by such
deeply bound trimers whose size is set by the thickness of
layers.
V. CONCLUSIONS
We found that the dimer Bose-Einstein condensation
< d and
appears in the strong coupling region where aeff ∼
< O(1), while the trimer Fermi gas appears in the
aeff kFA ∼
> d and kFA d ∼
< O(1). Assuming
unitarity region where |aeff | ∼
that the rest of the phase diagram is occupied by the interlayer
s-wave or intralayer p-wave superfluidity found in the weak
coupling region, we conclude that the system has the rich
phase diagram shown in Fig. 1. Our results can be tested by
using a ultracold Fermi-Fermi mixture of 6 Li and 40 K with
a species-selective optical lattice. How the above four phases
meet at the center of the phase diagram (shaded region in Fig. 1)
and whether more quantum phases appear in our system are
interesting open questions and can be addressed, in principle,
by future experiments. In particular, further study on how the
dilute gas of trimers evolves into a gas of their constituents as
kFA d is increased may shed light on a similar problem of the
transition from nuclear matter to quark matter existing in the
core of neutron stars. Finally, we emphasize that the scheme
to induce long-range interlayer correlations in ultracold atoms
that is presented in this article by taking the bilayer Fermi gas as
an example can be widely extended to multilayer geometries,
multiwire geometries, Bose-Bose mixtures, and Bose-Fermi
mixtures.
ACKNOWLEDGMENTS
Y. N. is supported by MIT Pappalardo Fellowship in Physics
and the US Department of Energy Office of Nuclear Physics
under Grant No. DE-FG02-94ER40818.
[10] M. Taglieber, A.-C. Voigt, T. Aoki, T. W. Hänsch, and
K. Dieckmann, Phys. Rev. Lett. 100, 010401 (2008).
[11] E. Wille et al., Phys. Rev. Lett. 100, 053201 (2008).
[12] A.-C. Voigt et al., Phys. Rev. Lett. 102, 020405 (2009).
[13] V. Efimov, Phys. Lett. B 33, 563 (1970); Nucl. Phys. A 210, 157
(1973).
[14] Y. Nishida and S. Tan, Phys. Rev. Lett. 101, 170401 (2008).
[15] Y. Nishida, Ann. Phys. 324, 897 (2009).
[16] M. A. Ruderman and C. Kittel, Phys. Rev. 96, 99 (1954).
[17] T. Kasuya, Prog. Theor. Phys. 16, 45 (1956).
[18] K. Yosida, Phys. Rev. 106, 893 (1957).
[19] N. Read and D. Green, Phys. Rev. B 61, 10267 (2000).
[20] S. Tewari, S. Das Sarma, C. Nayak, C. Zhang, and P. Zoller,
Phys. Rev. Lett. 98, 010506 (2007).
[21] N. Prokof’ev, O. Ruebenacker, and B. Svistunov, Phys. Rev.
Lett. 87, 270402 (2001); N. Prokof’ev and B. Svistunov, Phys.
Rev. A 66, 043608 (2002).
011605-4
Download